Understanding the mechanisms of protein folding and misfolding and accurately predicting protein structures are two of the important challenges facing the scientific community. Detailed knowledge of the molecular events leading to the formation of both native and non-native states are the basis for a full elucidation of the protein folding mechanisms. With the rapid progress facilitated by the high-throughput structural characterization of representative protein sequence families, an essential need is the highly accurate computational methods that can reliably generate near-experimental quality structural models for capitalizing on the investment in the structural genomics. Availability of such methods would enable accurate modeling of protein structures which would have significant impact on a range of fields including biotechnology, pharmaceutical industry, drug discovery, and life sciences in general. Duan and Zhou propose to combine the strengths of the groups with complementary expertise to develop computational methods for protein structure modeling and refinement with the ultimate goal to produce highly accurate and reliable methods for protein structure prediction that have comparable accuracy to experimental techniques. Aim 1: Duan and Zhou propose to develop novel conformational sampling method for protein structure refinement in the first specific aim. A recently developed "Grow-to-Fit" method will be utilized and further developed to enable accurate identification of the near-native structures from a large ensemble of perspective protein structures. Further development of the methods will facilitate structural refinement which will help to improve the structures to be comparably accurate as those obtained from experimental techniques. Aim 2: Duan and Zhou propose to develop effective free energy (scoring) functions for accurate all-atom modeling of protein structures. This novel scoring function is based on the synergistic concept of integrating both knowledge-based statistical potential and the all-atom physics-based force field. Furthermore, comparison to the statistical potential will allow critical assessment of the force field parameters and solvation models. Aim 3: Duan and Zhou propose to examine the roles of protein native structure topology in protein folding using FSD1, Protein G and Protein L and their respective topologically distinct mutants as the model systems;to study the dependence of tertiary structure formation on secondary structures. Comparison with experiments, including direct tests on the predictive ability of our model will be an integral part of our study and will be instrumental for a close scrutiny on the approach. PUBLIC HEALTH RELEVANCE: To understand the basic rules of life, how cell works, it is necessary to know the protein structures that are critically important to understand how they work. This proposal is motivated by the need to develop computational method to reliably predict protein structures based on the primary sequence. Because protein structures are also enormously useful in drug discovery, a potential impact of the proposed work in human health is in the area of development of novel therapeutics.